Residual life estimation based on a generalized Wiener degradation process

The issue of residual life (RL) estimation, which has been widely studied recently, plays an important role in scheduling maintenance. In this work, we present an adaptive method of RL estimation based on a generalized Wiener degradation process which subsumes several existing models as limiting cases. The nonlinearity, the temporal uncertainty, and the product-to-product variability of the degradation are jointly taken into account in the proposed degradation model. Under a mild assumption, an analytical approximation to the probability density function of the RL is derived in a closed-form, which becomes quite useful in maintenance decision making. The unknown parameters of the model that characterize the population-based degradation characteristics are obtained by using the maximum likelihood approach, while the parameters that describe the online product-specific characteristic are estimated by using the Markov chain Monte Carlo (MCMC) method. Once new degradation data information of the target product becomes available, the degradation model is first updated based on the degradation history up to the current time through a strong tracking filter, and then the RL is estimated sequentially. In this way, the RL of a product can be estimated in an adaptive manner. Finally, the validity of the proposed method is demonstrated with an illustrative example concerning fatigue cracks.